Research
Adaptive accelerated optimistic dual averaging method with p-norm preconditioning
The paper introduces A²ODA, an adaptive accelerated optimistic dual averaging method for unconstrained convex optimization that uses low-cost preconditioning based on p-norm geometry for p in [2,∞]. The method is line-search-free and uses a new stepsize selection based on an aggregated affine lower model, with a simple unified update structure across all p. Under generalized local smoothness, it derives a convergence bound separating constant and gap-dependent curvature, recovering a nearly optimal rate in the local smoothness case.
first-order optimizationadaptive methodsconvergence analysis